⚠ Draft RFC-backed issue.
Scope and architecture still under active review.
Do not implement without explicit validation.
Problem / Need
Define the execution invariants that make ruby-dag a reliable DAG execution kernel: deterministic runs, idempotent node execution, replayability, and clear boundaries between pure computation and effects.
Proposed scope
- Deterministic execution hashing
- Replayable scheduler
- Node idempotency contract
- Execution trace model
Acceptance criteria
- Same DAG definition + same input produce the same logical result hash.
- Re-running an idempotent node does not duplicate observable logical output.
- A run can be replayed and reconstruct the same state sequence.
- The kernel exposes calculation/effect boundaries without depending on LLM or host-specific code.
Draft sub-issues
Review notes
Move this issue from DRAFT/RFC to BACKLOG only after the architecture is accepted.
Problem / Need
Define the execution invariants that make
ruby-daga reliable DAG execution kernel: deterministic runs, idempotent node execution, replayability, and clear boundaries between pure computation and effects.Proposed scope
Acceptance criteria
Draft sub-issues
Review notes
Move this issue from
DRAFT/RFCtoBACKLOGonly after the architecture is accepted.